tackling the scalability challenge in artificial ......sustaining this benefit is the difficulty...
TRANSCRIPT
TACKLING THESCALABILITY CHALLENGEIN ARTIFICIAL INTELLIGENCEFOR CUSTOMER VALUE MANAGEMENT
Artificial Intelligence (ai) Is Witnessing Large-scale Implementations In Several Enterprise Functions3
5
810
How AI and ML Could Revolutionize Customer
Value Management
The State of AI Implementations TodayChallenges
Holding Back Organizations from
Scaling AI
Best Practices For A Sustainable Ai Strategy
13
C O N T E N T
2
ARTIFICIAL INTELLIGENCE (AI) IS WITNESSING LARGE-SCALE IMPLEMENTATIONS IN SEVERAL ENTERPRISE FUNCTIONS
According to PwC,
16 per cent of companies have
implemented AI pilots in
discrete areas, while 15 per
cent are planning expansion
into multiple areas; another 27
per cent of companies already
have AI projects in action
across different areas. And one
of the prime transformation
candidates for AI is customer
engagement and how to
maximize its value.
3
Mckinsey finds that stakeholders in marketing and sales are most likely
to achieve revenue gains from AI, far ahead of other functions like supply
chains, risk management, and HR. But one of the key challenges of
sustaining this benefit is the difficulty of scaling AI beyond a certain point.
To effectively use AI (and related technologies like NLP, OCR, ML, Neural
Networks and Deep Learning and augmented analytics) in customer
value management, it’s vital that organizations overcome technology
hurdles in scaling AI capabilities. The impacts of business processes
and the organization’s people assets also require careful investigation.
of companies have implemented AI pilots16%
are planning expansion into multiple areas15%
of companies already have AI projects in action27%
https://www.mckinsey.com/~/media/McKinsey/Featured%20Insights/Artificial%20Intelligence/Global%20AI%20Survey%20AI%20proves%20its%20worth%20but%20few%20scale%20impact/Global-AI-Survey-AI-proves-its-worth-but-few-scale-impact.ashx
https://www.pwc.com/us/en/services/consulting/library/artificial-intelligence-predictions-2019.html1
2
4
HOW AI AND ML COULD REVOLUTIONIZE CUSTOMER VALUE MANAGEMENT
The modern customer value management (CVM) function relies heavily
on data to gauge customer expectations, anticipate demand, and
preemptively design products/offerings to stay a step ahead of the
competition. In telecom, this is particularly important as customers are
eager to interact with service providers across a variety of channels.
Engagement analytics from these channels – be it social media, a carrier’s
mobile application, non carrier applications, such as travel applications,
hotel applications, food delivery applications, a branded website, affiliate
websites, Chatbots or the humble email – could go a long way in garnering
an accurate 360-degree view of the customer with versatile persona view
of the customer
5
This analytics data, combined with Machine Learning (ML), can power
more accurate decisions on the customer journey. For example, ML can
be leveraged to identify opportunities for making product recommendations,
including when not to make a recommendation offer. A keen understanding
of customer behavior would help AI engines target customers in a more
targeted way.
In fact, AI takes the power of data analytics to a whole new level. ML is a
specialized AI technique that helps analytics engines to learn from
historical data and past behavioral patterns. It can observe frequently
used channels and the most popular interaction times for a customer and
recommend when and how a carrier should approach the customer for
maximum engagement. This significantly improves upselling/cross-selling
success rates. Another component of AI which is Deep Learning and
Neural Networks helps to trap real time customer behaviour and suggest
real time recommendation based on very recent and new customer
behaviours which are not there in the history. This makes AI more suitable
with the changing impacting factors such as environments, weathers,
emergency conditions, competitions offerings, Geofencing etc.
6
All of these elements together drive up customer lifetime value by empowering carriers to interact with them more meaningfully. It is estimated that frequent engagement can increase revenue per customer by as much as 40 per cent; personalizing customer experiences based on Ai insights can boost ROI by 5-6X times.
Another way in which AI transforms customer value management (CVM) is via its ability to convert unstructured data into meaningful insights. Optical character recognition (OCR) extracts alphanumeric inputs from images, screenshots, PDFs, and other unstructured data assets. This can be auto-populated into the customer database without any intervention from a contact center executive. As a result, carriers can maintain a comprehensive and dynamically updated customer database, at optimized effort levels. The AI driven audio to text convertors help to understand customer sentiments on a audio call to Contact Center and can be helpful for generation of automated and more complete Net Promoter Scores. This is further amplified via natural language processing/ generation (NLP/NLG), which can translate information in natural languages like English into a machine-readable format. This plays a major role in social media analytics for telecom, scouring social platforms to find brand mentions, negative comments, positive feedback, and other sentiment indicators.
Owing to these benefits, carriers are eager to invest in AI deployments, targeting specific use cases like sentiment analysis on social media ,predictive email marketing automation, Audio Sentiment Scoring , Geofencing Offer Recommendations . Unfortunately, these projects often struggle to scale up in tandem with business growth, even as organizations find it challenging to integrate discrete AI pilots with the overall enterprise digital footprint.
7
Analytics data, combined with Machine Learning (ML), can power more accurate decision on the customer journey.
THE STATE OF AI IMPLEMENTATIONS TODAY
As AI fast becomes a mainstream technology (following in the heels of
once-disruptive trends such as the cloud and big data), issues around
scalability begin to emerge. Accenture found that 84 per cent of C-level
leaders believe scaling AI is integral for the business strategies, but only
16 per cent have successfully gone beyond early-stage experimentation.
Worryingly, 75 per cent said that they risk going out of business in the
next five years if the scalability issue in AI isn’t addressed.
8
Clearly, AI is yet to find widespread adoption and continues to be limited
to sporadic CVM use cases. This isn’t due to lack of interest, given that
one out of five companies (across industries) are planning industry-wide
AI deployment, moving ahead of isolated pilots . This brings us to a key
question: why are organizations, then, not implementing AI at scale?
Zeroing in on the telecom sector, research suggests that AI implementation
is yet to reach its promised potential. Only 41 per cent of operators use
advanced analytics (including AI) to onboard new customers; only half
are using this technology to auto-match customers with the right subscription
plans. And even in today’s digital world, 94 per cent of operators
communicate with their customers via SMS – even more worryingly, texts
are personalized in only 20 per cent of cases .
9
CHALLENGES HOLDING BACK ORGANIZATIONS FROM SCALING AI
Setting aside this foundational challenge, carriers often find it difficult to
train their AI engines with the requisite data sets and learning techniques.
In the absence of reliable, comprehensive customer data and effective
feedback from the trainer, AI engines will soon fall short of expected
performance benchmarks in more grueling interaction scenarios.
There are several reasons why carriers could fail to deploy AI across the
entire CVM function, in spite of successful pilots. To begin with, there are
technical challenges that limit AI ‘s capabilities. If the objective of AI is
to mimic human-level intelligence and understanding (so that a chatbot
could hypothetically be as sensitive and versatile as a contact center
executive), there is a long way to go. A human being can spot the
difference between a massive variety of objects – between 1022 and
1048 objects, according to research . But the most powerful AI engines
today perform approximately 1020 times worse than human cognition.
10
Another bottleneck in large-scale AI implementation is the skills level in
a telecom organization. Developing, training, managing, and utilizing Ai
requires highly-specific skill sets, over and above analytics capabilities.
PwC found that nearly one out of 10 companies is yet to develop any
sort of plan for building an AI-ready workforce, despite acknowledging
its need . In contrast, strategic scalers of AI (as identified by Accenture)
are 1.5X times more likely to receive formal training, 2X more likely to
understand AI’s application in the job role, and 1.7X times more likely to
know how to implement AI responsibly than laggard organizations.
Finally, one of the biggest roadblocks to scaling AI is the absence of a
unified strategy and leadership vision. In most cases, there is no
centralized owner for AI, which means that non-technical employees in
the CVM function could be tasked with AI success. In many cases, AI
falls entirely under technical teams, impeding collaboration with marketing,
customer support and CVM. According to PwC, the data and analytics
team owns AI in approximately 1 out of 5 companies; for 14%, the
automation team owns AI. Only 15% have an enterprise-wide AI leader
to guide implementations in a single cohesive direction.
11
While the first challenge may require technology advancements at the
foundation, skill gaps, fragmented strategies, and the lack of clear
ownership can be effectively addressed to accelerate AI growth.
Investment needed on AI tools and technologies to get the scalable,
user friendly, lower lead times benefits from the technology for solving
specific business issues. Where AI needs to handle unstructured data ,
the investment is needed in Hadoop Clusters that have cost involved of
enterprise licenses. Understanding AI by the business to think in terms
of solving the business case is another hurdle. The investments of time
and money in training CxOs to ground level staff is huge to start using AI
for the Business problem solutions.
https://www.pwc.com/us/en/services/consulting/library/artificial-intelligence-predictions-2019.html
https://newsroom.accenture.com/news/failure-to-scale-artificial-intelligence-could-put-75-percent-of-organizations-out-of-business-accenture-study-shows.htm
https://arxiv.org/abs/1505.00775
https://www.pwc.com/us/en/services/consulting/library/artificial-intelligence-predictions-2019.html
https://www.prnewswire.com/news-releases/global-survey-reveals-telecommunications-providers-lag-in-use-of-applied-ai-and-machine-learning-300827892.html5
6
7
8
9
12
BEST PRACTICES FOR A SUSTAINABLE AI STRATEGYTo ensure that investments in AI for increasing customer value return the
expected ROI in the long-term, organizations can:
The strategic positions like Chief Data Science Officer or
Chief AI Officers are required to focus on creating AI driven
initiatives across organisation. This will provide good thrust on
investing in AI and get ROI from the same across multiple
functions of the business. The focus on an AI-driven roadmap
for three to five years will be enabled with this.
13
Strengthen the underlying bedrock of data - By collecting data
regularly from omnichannel sources, carriers can train the AI
engine to better understand the customer and intercept queries.
This requires a data lake infrastructure that is equipped to handle
big data and a governance framework to spot data gaps.
Implement a workbench for data scientists - A workbench will
allow employees to gain self-service access to data, thereby
simplifying the ML prototyping process and improving
productivity for every stakeholder. The more advanced, intuitive
and user friendly workbench tools are required so that business
analysts and business teams can start using the same to solve
the simple to medium-complex business problems. Even data
scientists can leverage modeling workbenches to speed up
AI implementation exponentally.
Leverage AIOps – Advanced AIOps can identify AI use cases
to transform IT and network capabilities for greater service
availability, optimized internal processes, and better customer
experiences. Beyond IT operations, AIOps will also influence
CVM finds a survey by TM Forum, according to which customer
experience enhancement (77%), OpEx reduction via automation
and closed loop systems (62%), and predicting IT outages/
failures (45%) were the top three drivers for AI adoption among
communication service providers.
14
Reimagine the culture - This is a critical cog for scalable AI
implementation. Companies need to step away from a single
leader-driven decision-making process and embrace data-
driven strategies; this would mean large-scale democratization
of the workforce, bolstered by adequate upskilling.
Test for bias - Bias is a massive concern for AI implementation,
particularly in CVM, as biased customer data could skew the
insights generated by AI. The testing process must factor in
this issue, ensuring that there is no risk of unethical AI or
discriminatory data insights that adversely impact both the
customer and business success.
Use of AI for telecom companies in the current Emergency
Global Pandemic situations and post pandemic situations can
be enhanced across network capacity and fault predictions,
big data analytics based applications for contact tracing, et
all. Other use cases include enhacing cybersecurity by detecting
security threat over telecom networks predictively, better AI
power customer insights for CRM and retail stores in the remote
contacting scenarios and aligning the network elements
components supply chain before the failure happens to avoid
network down times.
https://inform.tmforum.org/catalyst/2019/04/repeat-aiops-future-telcos/
https://www.mckinsey.com/~/media/McKinsey/Business%20Functions/McKinsey%20Digital/Our%20Insights/Driving%20impact%20at%20scale%20from%20automation%20and%20AI/Driving-impact-at-scale-from-automation-and-AI.ashx
11
12
15
CONCLUSION – THE ROAD AHEADIt is estimated that AI could have over a $100 billion impact on the
telecom sector, if implemented effectively . To realize this potential, it is
vital to overcome the scalability hurdle and embed AI into the overall
enterprise fabric, impacting every customer interaction, engagement, and
offering. The potential of AI will only increase, as computing technology
becomes more powerful, affordable, and accessible.
By implementing the aforementioned best practices, telecom companies
can build a launchpad for scalable AI, dramatically increasing customer
value. Better customer experience is among the top five monetizable
benefits of AI – to take advantage of this, a robust data governance
structure is required, in conjunction with AI-aligned leadership strategies
and the requisite employee skill sets.
For more information, please visit www.comviva.com
Comviva is the global leader of mobility solutions and a part of the $21 billion Mahindra Group. With customer centricity, innovation and ethical corporate governance at its core, the company’s offerings are broadly divided into three categories-Financial Solutions, Digital Systems and Growth Marketing. Its extensive portfolio of solutions spans digital financial services, customer value management, messaging and broadband solution and digital lifestyle services. The company strives to enable service providers to enhance customer experience, resolve real, on-ground challenges and leverage technology to transform the lives of customers. Comviva’s solutions are deployed by over 130 mobile service providers and financial institutions in over 95 countries and enrich the lives of over two billion people to deliver a better future.
All trade marks, trade names, symbols, images, and contents etc. used in this
document are the proprietary information of Comviva Technologies Limited.
Unauthorized copying and distribution is prohibited.
©2019 Comviva Technologies Limited. All Rights Reserved.
ABOUT COMVIVA